Journal of Child Psychology and Psychiatry 45:2 (2004), pp 212–229

Adult outcome for children with autism

Patricia Howlin,1 Susan Goode,2 Jane Hutton,3 and Michael Rutter4
1
St George’s Hospital Medical School, London, UK; 2Croydon CAMHS South London and Maudsley Trust, UK;
3
South London & Maudsley NHS Trust, UK; 4Institute of Psychiatry, London, UK

Background: Information on long-term prognosis in autism is limited. Outcome is known to be poor for
those with an IQ below 50, but there have been few systematic studies of individuals with an IQ above
this. Method: Sixty-eight individuals meeting criteria for autism and with a performance IQ of 50 or
above in childhood were followed up as adults. Their mean age when first seen was 7 years (range 3–15
years); at follow-up the average age was 29 years (range 21–48 years). Outcome measures included
standardised cognitive, language and attainment tests. Information on social, communication and
behavioural problems was obtained from the Autism Diagnostic Interview (ADI). Results: Although a
minority of adults had achieved relatively high levels of independence, most remained very dependent
on their families or other support services. Few lived alone, had close friends, or permanent employ-
ment. Communication generally was impaired, and reading and spelling abilities were poor. Stereotyped
behaviours or interests frequently persisted into adulthood. Ten individuals had developed epilepsy.
Overall, only 12% were rated as having a ÔVery GoodÕ outcome; 10% were rated as ÔGoodÕ and 19% as
ÔFairÕ. The majority was rated as having a ÔPoorÕ (46%) or ÔVery PoorÕ (12%) outcome. Individuals with a
childhood performance IQ of at least 70 had a significantly better outcome than those with an IQ below
this. However, within the normal IQ range outcome was very variable and, on an individual level, neither
verbal nor performance IQ proved to be consistent prognostic indicators. Conclusions: Although
outcome for adults with autism has improved over recent years, many remain highly dependent on
others for support. This study provides some information on prognostic indicators, but more fine-
grained research is needed into the childhood variables that are associated with good or poor out-
come. Keywords: Autistic disorder, prognosis, adulthood, follow-up studies. Abbreviation: ADI:
Autism Diagnostic Interview.

individuals had jobs and one was still at college.

Studies of autism in adult life
For the parents of any child with a developmental
disorder, accurate information about future progress
can be crucial in helping them come to terms with
the diagnosis and in assisting them to develop ef-
fective ways of coping. Predicting outcome in autism,
however, presents particular problems because of
the very wide spectrum of cognitive, linguistic, social
and behavioural functioning associated with the
condition. Moreover, despite a number of follow-up
studies of older adolescents or adults with autism,
information about prognosis remains inconclusive
because of the heterogeneity of subjects, the vari-
ability of outcome measures used, and other funda-
mental problems of research design (see Howlin &
Goode, 1998).
The earliest follow-up studies were largely anec-
dotal and involved people of very different ages, so that
it was difficult to determine the outcome for adults
specifically (e.g., Creak, 1963; Eisenberg, 1956).
However, in 1973, Kanner reported on 96 individuals,
mostly in their twenties and thirties. The majority was
highly dependent, living with parents, in sheltered
communities, in state institutions for people with
learning disabilities or in psychiatric hospitals. Out-
come was more positive for those with better-devel-
oped communication skills and amongst this group
just over half was functioning relatively well. Eleven

Association for Child Psychology and Psychiatry, 2004.
Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA
Seven had their own homes and one man (a suc-
cessful music composer) was married with a child.
The first systematic outcome studies were conduc-
ted by Rutter and his colleagues (Lockyer & Rutter,
1969, 1970; Rutter, Greenfeld, & Lockyer, 1967;
Rutter & Lockyer, 1967). Thirty-eight individuals
were aged 16 years or older, having initially been
diagnosed in the 1950s and early 1960s. At follow-up
over half were in long-stay hospitals; 7 lived with their
parents, with no outside occupation, 4 others atten-
ded day centres and 3 were living in residential com-
munities; only 3 had paid jobs. Overall, 14% were said
to have made a ÔGoodÕ social adjustment; 25% were
rated as ÔFairÕ and 61% as ÔPoorÕ.
Lotter (1974a, b) followed up 29 individuals, aged
16 to 18 years, who had been diagnosed as autistic
when younger. Amongst the 22 who had left school,
only one had a job and almost half were in long-
stay hospital provision. Two individuals were living
at home and 5 were attending day training centres.
Fourteen per cent were rated as having a ÔGoodÕ out-
come; 23% as ÔFairÕ, and 63% as ÔPoorÕ or ÔVery PoorÕ.
In 1987, Gillberg and Steffenburg reported on a
group of 23 individuals aged 16 or over, living in
Sweden. Only one person was found to be fully self-
supporting; of the remainder, 48% were rated as
having a ÔFairÕ outcome and the same proportion as
ÔPoorÕ/ÔVery PoorÕ.

Kobayashi, Murata, and Yashinaga and col-
leagues (1992) used a postal survey to follow up
201 people, aged 18 to 33 years, in Japan. The
average follow-up period was fifteen years. Four
individuals, all male, had died; 5% were still at-
tending school or college; 20% were employed,
mostly in food and service industries. All but 3 of
those with jobs still lived with their parents; one
was in a group home and 2 had their own apart-
ments; none was married. Twenty-seven per cent of
the group was rated as having a ÔGoodÕ or ÔVery
GoodÕ outcome, on the basis of their adaptive skills.
The same proportion was rated as ÔFairÕ, whilst 46%
were rated as ÔPoorÕ or ÔVery PoorÕ.
A telephone survey by Ballaban-Gil, Rapin, Tuch-
man, and Shinnar (1996) found that amongst 45
adults, initially diagnosed as children, over half
(53%) were in residential placements and only one
was living independently. Three people had died.
Eleven per cent were in regular employment (all in
menial jobs) and a further 16% were in sheltered
placements. Rates of behavioural difficulties were
high and only 3 adults were rated as having no social
deficits. Estimates of changes in IQ levels from child
to adulthood indicated that cognitive functioning
remained very stable and only 18% showed a marked
change in IQ. When change did occur, this usually
indicated an improvement rather than decline in
intellectual ability.
Comparisons between studies need to be treated
with caution because of differences in sample se-
lection and in the measures used. Most investiga-
tions have involved relatively small groups of
subjects, diagnostic criteria are sometimes imprecise
and/or the quality of data on early intellectual
functioning is poor. Overall judgements of whether
outcome is ÔGoodÕ, ÔFairÕ or ÔPoorÕ also tend to be
based on variable criteria. Nevertheless, the one
consistent conclusion from these reports is that a
positive prognosis in adulthood is highly dependent
on IQ – very few individuals with an IQ below 50
achieve good functioning in adulthood (Gillberg &
Steffenburg, 1987; Lockyer & Rutter, 1969; Lord &
Bailey, 2002; Lotter, 1974b). Outcome is also poor
for those who fail to acquire some useful language by
the age of 6 years or so. However, even amongst
individuals of higher IQ outcome is very variable.
Rumsey, Rapoport, and Sceery (1985) followed up
14 men aged between 18 and 39 years of age, all of
whom fulfilled DSM-III criteria for autism. Verbal
and/or Performance IQ scores were 60 or above.
Socially, all the group continued to have marked
difficulties. Only one individual had friends and four
people were in independent employment. Most (71%)
remained very dependent on their parents or others
for support and one was in a state hospital. Similarly
low levels of employment and social functioning were
reported by Tantam (1991) in his cross-sectional
study of 46 individuals (mean age 24 years) with
Asperger syndrome.
Adult outcome of children with autism 213
Venter, Lord, and Schopler (1992) described out-
come for 22 individuals aged 18 years or over who
had a pre-school IQ of 60+. Around a third were
competitively employed, but again jobs were gener-
ally at a very low level and the majority was in shel-
tered employment or special training programmes;
3 had no occupation. Only 4 individuals (18%) lived
more or less independently. However, Szatmari
and his colleagues (Szatmari, Bartolucci, Bremner,
Bond, & Rich, 1989) reported rather more positive
findings for their group of 12 males and 4 females (all
17 years or over; mean IQ >90). Half had attended
college or university, and over a third were in regu-
lar, fulltime employment. Half were described as
being completely independent, although some of
these still lived at home. Although over half the
group had never formed close relationships, a
quarter had dated regularly or had long-term rela-
tionships and one was married.
More recently, Mawhood and colleagues (Howlin,
Mawhood, & Rutter, 2000; Mawhood, Howlin, &
Rutter, 2000) followed up 19 men with autism (mean
WAIS PIQ 83) who had initially been diagnosed
between 4 and 9 years of age. Although the majority
had improved over time, all showed continuing prob-
lems in communication, social relationships and
independence. Almost half remained socially isol-
ated, only 3 lived independently and over two-thirds
had significant difficulties associated with obses-
sional or ritualistic tendencies. Only 3 individuals
(16%) were considered to have a good outcome;
2 (10%) remained moderately impaired and 14 (74%)
continued to show substantial impairments.
Background to the present study
Previous follow-up studies have generally involved
only small numbers of individuals above the age of
21; ratings of adult functioning have frequently
been somewhat unsystematic, and the IQ range has
either been very extensive (ranging from severe
retardation to average) or restricted to high func-
tioning participants only. Studies involving mixed
IQ groups have generally found that outcome is
poorest in individuals with an IQ below 50 (see Lord
& Bailey, 2002). However, few people with this
degree of cognitive impairment, whether or not they
are autistic, achieve high levels of independence as
adults. An investigation of factors influencing long-
term outcome in individuals with autism with an IQ
above 50 is, therefore, of considerable practical
importance.
In the present study, cases were only included if
they had a non-verbal IQ in childhood of 50 or above.
Follow-up data included detailed psychometric test-
ing, and social functioning in many different areas
was assessed using standardised measures. The
longitudinal design also made it possible to explore
the relationship between IQ in childhood and adult
outcome variables.
214 Patricia Howlin et al.
The paper focuses on 3 main questions:
• What is the long-term outcome, in terms of social,
cognitive, linguistic and behavioural functioning
for individuals with autism who, as children, had a
performance IQ of at least 50?
• How stable are childhood measures of IQ?
• How does early cognitive ability relate to prognosis
in adulthood and what other factors are predictive
of outcome?
Methods
Participants
The participants were referred to the Children’s De-
partment at the Maudsley Hospital, London, because of
concerns about their development. All were diagnosed
as having an autistic disorder. Diagnosis was based on
the criteria applicable at the time the children first at-
tended. Although classification systems have been
modified over the years, the 4 core criteria currently
used in DSM-IV-TR and ICD-10 remain closely similar
to those defined by Rutter (1966, 1972, 1978), which
were used for cases seen in the earlier years, viz:
• Onset prior to 30/36 months
• Impaired social development which has a number of
special characteristics and is out of keeping with the
child’s intellectual level
• Delayed and deviant language development, which
also has certain defined features and which, again, is
out of keeping with the child’s intellectual level
• ÔInsistence on samenessÕ as shown by stereotyped
play patterns, abnormal preoccupations or resistance
to change.
At the start of the follow-up all case notes were checked
again by MR to ensure that they met criteria for autism.
Cases were excluded if they clearly failed to meet these
criteria. A number of other exclusion criteria were also
employed (see Appendix 1 for details). Individuals were
then included in the adult follow-up study if:
a) they had been seen as children (i.e., prior to 16
years) primarily for a diagnostic assessment (those
initially referred for deterioration or psychiatric dis-
orders other than autism were not included but the
presence of comorbid disorders was not a reason for
exclusion if individuals were referred for a diagnosis
of possible autism) and
b) they had a non-verbal IQ in childhood of 50 or above
and
c) they had reached the age of 21 at the time of follow-up.
Seventy-nine individuals meeting these criteria were
identified. Families were traced by means of previous
addresses, NHS registers and personal contacts and
were then approached in writing to request their
cooperation. One individual had died of status epilep-
ticus; 2 families refused to take part; 6 cases could not
be traced or did not respond (after repeated attempts to
contact them); 2 families were not contacted due to an
administrative error. The current age and initial IQ of
the 10 individuals who refused or were not contacted
did not differ significantly from the study sample (mean
age at time of follow-up 33.5 years; sd 10.3 years; initial
PIQ 82.3; sd 11.44).
The final sample comprised 68 individuals, but only
partial follow-up data were available for one of these and
some ADI information was missing on another. The
average age of the group at follow-up was 29.33 years
(sd 7.97 years; range 21.16 to 48.58 years). Nineteen
individuals had been first seen between 1950 and 1959;
28 between 1960 and 1969 and the remainder between
1970 and 1979. Over this period, and particularly dur-
ing the 1950s and 1960s, few children with autism were
seen at local clinics and hence the sample was much
less selective than would be the case today. Moreover,
findings from the Maudsley Hospital samples have been
consistently similar to those from comparable epidemi-
ological studies over the same period. The key feature of
all samples (clinical and epidemiological) seen many
years ago is that the prevailing diagnostic concept of
autism then was more restricted to seriously handicap-
ping disorders than would be the case today.
Characteristics of the group in childhood. The
group comprised 61 males and 7 females. This ratio is
higher than the male:female ratio of around 4:1 typic-
ally reported in the literature, probably because indi-
viduals with an IQ below 50 were excluded from the
sample and there tend to be more females within this
lower functioning group (Lord & Schopler, 1985). The
average age when children were initially seen was 7.24
years (sd 3.10 years; range 3.1 to 15.66 years).
Initial IQ estimates were based mainly on the Per-
formance Scale of the Wechsler Intelligence Test for
Children (WISC-R; Wechsler, 1974) or the Wechsler
Pre-School and Primary Scale of Intelligence (WPPSI;
Wechsler, 1990) (n ¼ 30), or scores on the Merrill Palmer
(Stutsman, 1948) (n ¼ 34). Two children were assessed
on the Leiter (Levine, 1982), one on the Stanford–Binet
(Terman & Merrill, 1961) and one on the Vineland
Adaptive Behavior Scales (Sparrow, Balla, & Cicchetti,
1984). The mean Performance IQ score derived from the
most detailed testing was 80.21 (sd 19.28, range 51–137).
On verbal tests 22 children obtained a verbal IQ score
on either the WISC or WPPSI, 10 scored on the Peabody
Picture Vocabulary Test (PPVT; Dunn, 1965) or British
Picture Vocabulary Scale (BPVS; Dunn, Dunn, Whetton,
& Pintillie, 1982), 4 on the Reynell Language Scales
(Reynell, 1977), 2 on the Mecham Scale (Mecham, 1958)
and one on the Stanford–Binet vocabulary test (Terman
& Merrill, 1961). The average verbal IQ estimate (based
on scaled scores where appropriate; LA/CA · 100 if not)
was 61.49 (sd 21.26, range 21–106). For children scor-
ing on both performance and verbal tests there was a low
but significant correlation between performance IQ
scores and verbal IQ estimates at this age (n ¼ 39,
r ¼ .34, p ¼ .04) but performance scores were signific-
antly higher (t ¼ 4.99, df 38, p < .001).
Measures
Assessment measures in adulthood. Follow-up
interviews and assessments were conducted between
1985 and 1991, in almost all cases by SG, but with a
few assessments undertaken by PH and MR.
Diagnostic and behavioural assessment. Diagnosis
in adulthood was re-confirmed for all subjects using the

Autism Diagnostic Interview (ADI; Le Couteur et al.,
1989), generally with parents as informants. In one
case parents had died and information was collected
from a sibling. The ADI was also used to obtain ratings
of current language functioning (both appropriate and
inappropriate use) and to assess the severity of stereo-
typed behaviours/interests (see below).
Social functioning. Current levels of social compet-
ence were assessed using the ADI and information
was also obtained from parents on educational back-
ground, employment and independence. Because the
range of language skills in the group was too wide for
comparability of self-reports, the outcome measures
used were based on a combination of detailed,
standardised interview schedules of demonstrated
reliability and validity, and psychometric assess-
ments. Whenever relevant, assessments were supple-
mented by reports from employers, or other
professionals, and medical records. From the parental
questionnaire, a composite outcome rating score was
derived by summing individual scores for Occupation,
Friendships and Independent Living (see Appendix 2).
A rating of 0 indicated a ÔVery GoodÕ outcome (i.e.,
subjects achieving a high level of independence);
1 ¼ ÔGoodÕ outcome (generally in work but requiring
some degree of support in daily living); 2 ÔFairÕ (has
some degree of independence, and although requires
support and supervision does not need specialist re-
sidential provision); 3 ¼ ÔPoorÕ (requiring special resid-
ential provision/high level of support) and 4 ¼ ÔVery
PoorÕ (needing high-level hospital care).
Autistic-type behaviours in adulthood. The severity
of stereotyped and ritualistic behaviours was assessed
using a composite rating on the ADI (Le Couteur et al.,
1989), which included scores for verbal and beha-
vioural rituals, unusual preoccupations and object at-
tachments, and motor mannerisms (see Appendix 2 for
details).
Cognitive and linguistic abilities. At follow-up, the
participants completed a battery of cognitive assess-
ments appropriate to their developmental level. Ideally,
a Performance IQ score was obtained from the Wechsler
Adult Intelligence Scale-Revised (WAIS-R; Wechsler,
1981). Individuals unable to obtain a performance score
on the WAIS-R were assessed (in order of preference) on
the Raven’s Progressive Matrices (Raven, 1976), the
Leiter (Levine, 1982) or Merrill Palmer Scales (Stuts-
man, 1948). Full-scale and Verbal IQ scores were also
calculated for those individuals able to complete the
WAIS-R. The British Picture Vocabulary Scale (Dunn
et al., 1982) was used to assess language comprehen-
sion. Other aspects of language usage in adulthood were
assessed using the ADI which provided ratings of overall
language competence and extent of abnormal language
usage in adulthood, age of first words and phrases, and
level of language development by the age of 5 years (see
Appendix 2 for details of all language measures).
Attainments in reading and spelling were assessed by
means of the Neale Test of Reading Ability (Neale, 1977)
and the Schonell Spelling test (Schonell & Schonell,
1960).
Adult outcome of children with autism 215
Statistical analysis
Parametric analyses were used for comparison where the
data permitted. When sub-group sizes were small and/or
variances significantly unequal, non-parametric tests,
with the appropriate corrections for ties or sample size,
were used. Non-parametric analyses were also used
when comparing overall ratings of abnormality/com-
petence (as on ADI data). However, in the tables means
rather than medians are reported as these illustrate more
clearly the differences between sub-groups.
Because of the relatively large number of compar-
isons conducted, the analysis focuses on the con-
sistency of the pattern of findings, rather than on the
statistical significance of individual variables. Single
results are not reported as significant unless they reach
a p value of at least .01; probability tests are two-tailed.
Results
Educational attainments
Only 10 individuals had been educated predomin-
antly in mainstream schools. Twenty-nine (43% of the
total group) had spent most of their school years in
specialist autistic provision; 9 (13%) had attended
schools for children with more general learning dis-
abilities and 16 (24%) had spent the majority of their
time in a variety of other educational settings such as
hospital schools, schools for emotional and beha-
vioural problems, language units, home tuition etc.
Data on schooling were not available for 4 individuals.
Forty-two people (62%) had spent at least some time in
boarding school. For most, school placements were
relatively stable with fewer than 3 changes of school
during either primary or senior school.
Academic attainments generally were not particu-
larly high, with over one-third of the group failing to
score on any of the attainment tests. Mean age
equivalent scores (and sd’s) on the Neale test of
Reading Accuracy and Comprehension, and on the
Schonell spelling test, were 10.54 years (2.13); 8.70
years (2.32) and 11.68 years (2.83) respectively.
Twenty-nine individuals scored at or below the basal
of 5–6 years on Accuracy, 32 were below basal for
Comprehension and 26 below basal for Spelling.
The majority (53; 78%) left school without any for-
mal qualifications. Of the 13 who did obtain academic
qualifications, 6 had obtained at least one GCSE or
equivalent and 2 had passed A levels; 2 had success-
fully completed diplomas (one in accounting, one in
design), 3 had obtained degrees (in science or com-
puting) and 2 of these had postgraduate qualifica-
tions.
Jobs
At the time of the follow-up (which, according to
government statistics, was a period of steadily in-
creasing employment in the UK), almost one-third of
the group (23 individuals) was in some form of
216 Patricia Howlin et al.

employment. Eight were working independently and
one man was a self-employed fabric printer, al-
though this did not provide him with a living wage.
Fourteen individuals worked on a supported/shel-
tered or voluntary basis (2 worked for the family
business and one in a shop run by his residential
centre). One other man, previously employed in a
factory, had been unemployed for some years. Be-
cause employment status was subject to change,
information on jobs was updated by Hutton (1998)
as part of a separate study. This showed no increase
in the numbers in independent work (see Table 1),
although 2 individuals who had experienced diffi-
culties in keeping jobs were now in supported job
schemes, but still on full salary. Of those initially in
sheltered schemes, one had moved into a supported
employment scheme for individuals capable of more
independent work; one had resigned his job and was
now unemployed; and another, who had been
working as a decorator for his father, no longer did
so. One man who was previously unemployed, and
another, initially working in a sheltered job scheme
as a grave digger, were now in specialist schemes run
by their day or residential centres. Ten further indi-
viduals who had previously had no jobs were in-
volved in off-site schemes organised by their day or
residential centres. Five other individuals carried out
some routine work (weaving, gardening, bakery, etc.)
within the residential centre. The remaining 27 were
described as being in general work/leisure pro-
grammes within their day or residential units. As is
evident from Table 1, most of the jobs were low level
and several had been found via parental contacts,
rather than through the open job market; pay was
generally poor. The majority was in sheltered
schemes or occupational programmes provided by
their residential or day centres. However, only one
person had no daytime occupation and this was due
to the breakdown of his previous residential place-
ment. (Information was missing for 2 cases.)
Friendships
Friendships were rated on the basis of parental
information from the ADI. ÔFrequencyÕ of friend-
ships was rated according to the numbers of

Table 1 Employment levels in adulthood ( fi indicates changes when cases reviewed in subsequent follow-up – Hutton, 1998)

Type of job Employment status

Scientific officer, oil company Independent

Electrical work Independent
Cartographer Independent
Postal assistant Independent
Factory work Independent
Factory work Independent
Computing Independent ( fi *supported)
Accounts Independent ( fi *supported)
Fabric design Self employed
Washing up Voluntary/low pay scheme
Grave digger Sheltered ( fi scheme run by residential centre)
Office/accounts assistant Sheltered ( fi *supported)
Factory work Sheltered ( fi resigned, now unemployed)
Charcoal burning/gardening Sheltered
Factory work Sheltered
Administrative assistant Sheltered
Data input Sheltered
Assembly work Sheltered
Supermarket trolleys Sheltered
Electronic work Sheltered
Special shop Sheltered ( fi factory assembly; scheme run by residential centre)
Decorating with dad Family based ( fi no job)
Office (with parents) Family based
Not initially in work (n ¼ 10) fi Scheme organised by residential/day centre
Part time office assistant
Wood/metal work · 2
Carpentry
Gardening · 2
Factory assembly work · 2
General work · 2
Not initially in jobs (n ¼ 5) fi Scheme within residential/day centre
Assembly work, gardening, weaving, bakery etc.
In day/residential unit but no specific work (n ¼ 27)
No occupation (n ¼ 1)
Not known (n ¼ 2)

*Supported schemes are those that offer assistance within the regular workplace. Job coaches fade out their support with time. The
level of support is much less than in sheltered schemes, and job levels are generally higher.

friends/acquaintances the individual had, and
ÔQualityÕ according to the degree of sharing and
mutual participation the relationship involved. In-
formation on friendships was missing for 2 indi-
viduals. Eighteen individuals (26%) were rated as
having a relationship with at least one other person
in their age group that involved participation in a
range of interests or activities, although in 2 cases
these interests were quite stereotyped. Ten (15%)
had some acquaintances with whom they might
talk or share activities but these were generally
within arranged social groups. The majority (38;
56%) was rated as having no friends or acquaint-
ances.
Thirteen individuals (19%) were considered by
their parents to have one or more relationships that
involved some degree of selectivity and sharing (of
enjoyment, activities or confidences) and self-
initiative in seeking contacts. Four more were des-
cribed as sharing activities with others but as
showing little pleasure in this; over half (35; 51%)
were said to have no friends with whom they shared
activities, and in 16 cases (23%) the quality of
friendships was unknown. Very few individuals had
had a close sexual relationship. At the time of the
follow-up one man had married, although he later
divorced. Two men have married subsequently.
Independent living
Three individuals lived by themselves, with only
limited parental support, and 4 others were in semi-
sheltered hostel-type accommodation, offering some
ongoing support. Twenty-six individuals, over a third
of the group, still lived at home (in 3 cases this was
because alternative placements could not be found
or had broken down). However, a few of those living
at home still had considerable independence; one,
for example, lived in a separate apartment at the end
of his parentsÕ garden and 8 others spent much of
their time away from home in independent social
activities. Half of the group was in some form of res-
idential accommodation – 14 in specialist autistic
provision with little independence; 12 in residential
settings with some degree of independence; and 8 in
long-stay hospital provision. In the hospital group, 4
individuals had extreme behavioural difficulties that
necessitated a very high level of supervisory care.
The others were all in the older age group (35+) and
had generally been placed in hospital accommoda-
tion because no alternatives had been available at
the time when they were seeking a placement. Data
were missing for one individual.
Overall social outcome
An overall estimate of social competence was derived
by summing the rating scores for ÔFriendshipsÕ, ÔWork
placementsÕ and ÔIndependent livingÕ. Information
Adult outcome of children with autism 217
was missing on one case but of the remainder, 8 in-
dividuals (12% of the sample) were rated as having a
ÔVery GoodÕ outcome, in that they were in paid em-
ployment, had some friends and a high level of inde-
pendence. A further 7 individuals were considered to
have a ÔGoodÕ outcome, i.e., they were working (albeit
with some support), could travel independently and
organise their own activities, and had some friends.
Thirteen (19%) had a ÔFairÕ outcome; they were gen-
erally living at home and required considerable sup-
port in daily living but they did have some degree of
autonomy. Thirty-one individuals (46%) were rated as
ÔPoorÕ – almost all of these were in residential accom-
modation with very limited autonomy or were living at
home because nowhere else would accept them. A
further 8 were in long-stay hospitals and were rated as
having a ÔVery PoorÕ outcome.
Autism-related problems
Autistic-type symptoms, including rituals and ste-
reotyped behaviours, resistance to change, unusual
object attachments and unusual/intense preoccu-
pations, were assessed using the ÔCurrentÕ section of
the ADI (Le Couteur et al., 1989). From this a com-
posite score for abnormalities related to ritualistic
and stereotyped behaviours was calculated (see Ap-
pendix 2). Information was missing on 2 cases. Of
the remainder, 8 individuals (12%) were rated as
showing few/no problems of this kind as adults, 28
(42%) showed mild problems, 23 (35%) moderate
problems and 7 (11%) severe problems.
Epilepsy
Ten individuals (15% of the total group) were
reported as having epilepsy (defined as at least one
fit after the age of 5 years). In 3 cases there had been
just one epileptic attack (although it was clear that
this was a major attack), 6 were on regular anti-
convulsant medication and one was undergoing
further testing to establish the most appropriate
drug to use. In all cases, the epilepsy was confirmed
by detailed interviewing on the nature and context of
the attacks.
Cognitive and language abilities
As adults, 44 individuals were able to obtain a Full
Scale IQ score on the WAIS-R; 2 others completed
the Performance scale only and one the Verbal scale
only. Of those unable to complete either of the
Wechsler scales, 15 were assessed on the Raven’s, 5
on the Leiter and one on the Merrill Palmer. Taking
scores from the highest-level IQ test completed (see
Appendix 2 for details), the overall mean perform-
ance IQ score was 75.00 (sd 21.52). The mean per-
formance IQ score of those tested on the WAIS-R was
84.48 (sd 16.06); for those tested on the Raven’s
it was 61.53 (sd 17.94) and for those who could
218 Patricia Howlin et al.

complete only the Merrill Palmer or Leiter it was
39.67 (sd 4.59).
Forty-five adults were testable on the WAIS-R
Verbal Scale. The mean verbal IQ of these indi-
viduals was 79.78 (sd 18.86). The difference between
WAIS Performance and Verbal IQ scores for those
completing both scales was not significant (t ¼ 1.97;
df 43, p ¼ .055). Sub-test profiles on the WAIS are
illustrated in Figure 1. Highest scores were for Block
Design, Object Assembly and Digit Span; the lowest
were for Comprehension, Picture Arrangement and
Picture Completion.
All adults were administered the British Picture
Vocabulary Scale but 4 individuals scored below
baseline (raw score £10) and for many of the re-
mainder it was not possible to assign a standard
score because their age was above the upper limit for
the test. Instead age equivalent scores were calcu-
lated. The average language age equivalent was 8.26
years (sd 6.21 years) but problems in calculating age
scores (see Appendix 2 for details) mean that this
estimate must be treated with considerable caution.
Nevertheless, the generally poor level of linguistic
functioning within the group is apparent from the
fact that 33 (48%) had a language age below 6 years;
24 (35%) scored within the 6 to 15 year range, and
only 11 (16%) scored above a 15-year level.
ADI ratings (Le Couteur et al., 1989) were used to
derive an overall composite score of language usage.
Seven individuals (10% of the group) were rated as
having good language; 21 (31%) as having mild im-
pairments; 6 (9%) as moderately impaired; 27 (40%)
as severely impaired; and 6 (9%) had no language.
Abnormal use of language was also rated using the
ADI. Fourteen individuals had no or insufficient
speech to be coded on this variable and information
was missing for one case. Twenty-three (43% of the
remainder) were rated as showing no or very few
12
10
Scaled score
abnormalities; 21 (40%) as showing occasional/mild
problems; and 9 (17%) as having moderate/severe
problems. (See Appendix 2 for full details of codings.)
I. Associations among adult outcome measures
Correlations among almost all outcome measures
were highly significant and substantial (see Table 2).
Thus individuals with the highest social outcome
ratings had higher scores on cognitive, language and
reading and spelling tests; they showed greater use
of social language and abnormal language features
and ritualistic behaviours were less frequent. Gen-
erally, correlations among social outcome measures
were higher with verbal IQ than with performance
IQ, although the differences were not significant.
The sample was also split into 3 sub-groups ac-
cording to overall outcome: Very/Good; Fair; Very/
Poor. One Way Analysis of Variance indicated signi-
ficant between group differences on all the principal
adult measures, although post-hoc (Scheffé) analy-
sis indicated that for many variables there was rel-
atively little difference between the Very/Good and
Fair Outcome groups (see Table 3).
II. The associations between childhood measures
and outcome in adulthood
A. Stability of IQ from childhood to adulthood
Performance IQ. Scores on tests of performance IQ
remained relatively stable over time. Although the
average time gap between initial and follow-up
assessments was 22.1 years (range 6.8 to 41.3
years), there was a significant correlation between
individualsÕ child and adult performance IQ (n ¼ 68,
r ¼ .54, p < .001; scores based on the most detailed
measure on each occasion). Thirty-one individuals
remained within their initial IQ band (i.e., ‡100;
70–99; 50–69) over time and 11 moved up one band.
Twenty-five individuals moved down by one band
and one showed a decrease of 2 bands (see Figure 2).
Stability tended to be greater amongst individuals
with an initial performance IQ of at least 70. Among

8 the 45 children scoring 70 to 100+ initially, 35 (78%)

remained in this range. In the lower IQ band there
6
4
Table 2 Correlations between social outcome and other vari-
2 ables in adulthood

0 Social outcome score

Si

Ar

Vo

Co

P
Di

PIQ .66 (n ¼ 67)

m an

Co

Ar
De

As
g

ith

ca

VIQ (or VIQ estimate) .81 (n ¼ 67)

it

il

r
s

m
s
b
sp

BPVS age equivalent .80 (n ¼ 67)

Reading comprehension .64 (n ¼ 35)
Subtest Reading accuracy .58 (n ¼ 38)
Spelling .51 (n ¼ 41)
Verbal scale Social use of language .67 (n ¼ 67)
Abnormal use of language ).45 (n ¼ 53)
Performance scale Ritualistic and stereotyped behaviours ).34 (n ¼ 66)

Figure 1 Profile of subtest scores on WAIS All correlations significant at p £ .01.

 Adult outcome of children with autism 219

Table 3 Cognitive and other adult scores according to overall outcome rating

1. 2. 3.
Very/Good Fair Very/Poor
Outcome Group1 fi (n ¼ 15) (n ¼ 13) (n ¼ 39)

Adult variable fl Mean (sd) Mean (sd) Mean (sd) F p Post–hoc (Scheffé)

PIQ 94.8 (12.2) 85.4 (14.3) 64.5 (19.5) 19.54 <.001 1¼ 2 > 3
VIQ 98.7 (16.9) 76.6 (11.9) 37.8 (26.9) 43.39 <.001 1> 2 > 3
BPVS age equivalent (yrs) 16.3 (4.5) 10.5 (5.0) 4.5 (3.3) 51.69 <.001 1¼ 2 > 3
Reading comprehension (yrs) 10.5 (2.2) 8.5 (2.3) 7.0 (0.5) 10.37 <.001 1¼ 2 > 3
Reading accuracy (yrs) 12.2 (1.1) 11.5 (1.7) 8.9 (1.9) 13.76 <.001 1¼ 2 > 3
Spelling (yrs) 13.1 (2.4) 12.5 (1.9) 9.7 (3.0) 7.55 .002 1¼ 2 > 3
Social outcome score2 2.4 (1.4) 6.2 (0.8) 9.4 (1.0) 237.07 <.001 1< 2 < 3
Social use of language2 3.6 (3.3) 7.4 (3.6) 14.7 (6.9) 23.10 <.001 1¼ 2 < 3
Abnormal use of language2 1.3 (1.4) 2.6 (2.9) 4.1 (2.7) 6.46 <.001 1¼ 2 < 3
Ritualistic and stereotyped behaviours2 3.3 (2.1) 5.2 (3.6) 7.3 (4.9) 5.47 <.001 1¼ 2 < 3
1
Outcome data missing on one case.
2
The higher the score the more abnormal the behaviour.

Verbal IQ. Just over half the sample (n ¼ 39) had

Init IQ FU.IQ obtained a score on a verbal test in childhood and
4
≥100 (n=10) ≥100 (7) within this group the correlation between child ver-
bal IQ (or estimates based on PPVT/BPVS) and adult
3 verbal IQ scores was significant (r ¼ .67; p < .001).
6
Init IQ 22 FU IQ However, as is apparent from Figure 3, there was
70-99 (35) 70-99 (36)
8 considerable movement in verbal IQ levels over time.
There was a slight overall rise in verbal IQ scores
9 (initial VIQ/VIQ estimate 61.49, sd 21.26; adult VIQ
Init IQ 5 FU IQ 69.64, sd 27.16; df 38, t ¼ 2.49, p ¼ .017). The
50-69 (23) 50-69 (14)
increase was greater for those scoring on Wechsler
10 tests on both occasions, from 70.64 (sd 15.26) in
1
childhood to 82.41 (sd 17.28) in adulthood (df 21,
FU IQ
t ¼ 4.0, p ¼ .001; child–adult VIQ r ¼ .65, n ¼ 22,
30-49 (11)
p ¼ .001). Stability was greatest in the 14 individu-
als who had initially achieved a verbal IQ (or IQ
equivalent) of ‡70 and all but 2 in this sub-group
had maintained the same level. Of the 23 initially
Figure 2 Stability of PIQ from child to adulthood scoring between 30 and 69, 16 (69%) had moved
upwards; only 4 remained the same and 3 had
was somewhat more movement, although there was moved downwards. Moreover, amongst the 31 indi-
no significant relationship between stability and viduals who had either been untestable, or scored
initial performance IQ (df 3,64; F ¼ 1.99, p ¼ .125).
Mean performance IQ scores from child to adult
testing fell slightly over time (initial PIQ 80.21, sd Init VIQ 12 FU. VIQ
19.28; adult PIQ 75.00, sd 21.52; df 67, t ¼ 2.19, ≥70 (n=14) ≥70 (32)
p ¼ .03). However, amongst those testable on 8
Wechsler tests on both occasions there was a small 3 1
increase in Performance IQ (from 81.84, sd 15.89, to Init VIQ 4 FU VIQ
50-69 (13) 5 50-69(14)
86.80, sd 16.68; t ¼ 1.48, df 24, p ¼ .15). Test–retest
correlations on the Wechsler tests were moderate 4
(child–adult WAIS-R PIQ r ¼ .47, n ¼ 25, p ¼ .02). Of
the 10 individuals whose performance IQ declined Init VIQ 9 1
30-49 (10) 1
from the average range initially (i.e., ‡70) to below 70
at follow-up, only 2 had initially been testable on the 2
WISC. In addition, all but 2 of the 10 were rated as
Init VIQ 18 FU VIQ
having very poor/no language when first tested and
<30/no test <30/no test
their language abilities in adulthood remained
(31) (22)
severely impaired. The one individual who showed
the largest decline (from 95 initially to only 34 at
follow-up) was testable only on the Leiter scales on
both occasions. Figure 3 Stability of VIQ from child to adulthood
220 Patricia Howlin et al.

Table 4a Initial cognitive level and outcome rating measures: differences between individuals with a childhood IQ ‡ 70 and IQ
50–69

Mean (sd)

Childhood IQ ‡70 (n ¼ 45a) 50–69 (n ¼ 23)

c
Parental rating Mann Whitney Zb p
Residential status 2.68 (1.52) 4.13 (1.79) 3.20 .001
Quality of friendships 1.75 (1.37) 2.69 (.70) 2.75 .006
Educational level 3.16 (1.43) 4.00 (0) 2.89 .004
Level of work 2.14 (1.17) 2.69 (.76) 2.13 .033
Total social outcome rating 6.38 (3.21) 8.83 (2.08) 3.12 .002
a
Parental data missing on 1 case in this group.
b
Means, rather than medians, are presented for ease of reference. However, because data were not normally distributed, non-
parametric statistics were used for group comparisons.
c
The higher the score the greater the abnormality, see Appendix 2.

Table 4b Initial cognitive level and outcome rating measures: differences between individuals with a childhood IQ ‡ 70 and
IQ 50–69

Childhood IQ ‡70 (n ¼ 45a) 50–69 (n ¼ 23)

c
Parental rating Mean (sd) Mean (sd) Mann Whitney Zb p
Social use of language 8.73 (5.89) 14.78 (8.52) 2.83 .005
Abnormal use of language 2.67 (2.82) 3.77 (2.20) 1.75 .080
Ritualistic behaviours 5.16 (3.33) 7.68 (5.78) 1.71 .087
Standard test scores Mean in years (sd) Mean in years (sd) t
BPVS age equivalent 10.31 (6.54) 4.24 (2.58) 5.54d <.001
Neale accuracy age 10.81 (2.13) 9.56 (1.91) 1.59 .118
Neale comprehension age 9.13 (2.44) 7.20 (.93) 3.41d .002
Spelling age 12.14 (2.75) 9.69 (2.43) 2.31 .026
a
Parental data missing on 1 case in each group.
b
Means for rating scores, rather than medians, are presented for ease of reference. However, because of the nature of these data,
non-parametric statistics were used for group comparisons.
c
The higher the score the greater the abnormality, see Appendix 2.
d
Equal variances not assumed.

<30 when first seen, over a third (13) were scoring at
a considerably higher level as adults – 9 above 70
and 4 between 50 and 69.
B. Childhood indicators of adult outcome
Performance IQ. When first assessed, all the chil-
dren had completed some form of non-verbal IQ test.
Thus it was possible to examine the impact of this
variable on later functioning in every case. Particip-
ants were split into 2 main IQ bands: those with a
childhood performance IQ in the normal range (i.e.,
70+, n ¼ 45; of whom 10 had an IQ ‡ 100) and those
with a performance IQ in the mildly intellectually
impaired range (i.e., 50–69; n ¼ 23).
There were significant differences between the 70+
and 50–69 groups on a number of social outcome
variables (see Table 4a). Of those with a performance
IQ initially below 70, only one was described as having
friends and only one was living semi-independently;
the remainder was rated as highly dependent. The 3
individuals who were living at home because no other
placement could be found for them, and 4 out of the 8
people in long-stay hospital care all had non-verbal
childhood IQs below 70 (no one with a childhood IQ
above 100 was in hospital care).
Although relatively few individuals had obtained
any formal qualifications, there was a significant
difference between the IQ bands. No one with an
initial performance IQ of below 70 had obtained any
formal qualifications and the childhood IQ scores of
the 5 individuals who went on to college or university
were all 70 or above (119, 97, 80, 76 and 70). Mean
ratings for levels of employment were similar but
almost all those currently in some form of employ-
ment (paid, voluntary or sheltered) had an initial
performance IQ of 70+; only one of the 9 people in
independent employment had an initial IQ below 70
(Fisher exact test for n in work in ‡70 and 50+
groups, p ¼ .005. See also Table 4a).
Group differences on the total social outcome
score (based on summed ratings for ÔFriendshipsÕ,
ÔWorkÕ and ÔIndependent livingÕ) also indicated a
significant difference between the two IQ groups (see
Table 4a). Of the 23 individuals with a non-verbal
childhood IQ below 70, only one achieved a ÔVery
GoodÕ outcome rating in adulthood; 3 others were
rated as ÔFairÕ; the remainder as ÔPoorÕ or ÔVery PoorÕ.
In the ‡70 group, 7 individuals were rated as ÔVery
GoodÕ, 7 as ÔGoodÕ, 10 as ÔFairÕ, and 20 as ÔPoorÕ/ÔVery
PoorÕ. Further examination of the data showed that

there was no difference in outcome between those
with an initial performance IQ of over 100 and those
with an initial IQ of 70–99 (Mann–Whitney Z ¼ 1.37,
p ¼ .17). Indeed, somewhat surprisingly, none of
those in the ÔVery GoodÕ outcome group as adults
had a childhood IQ of 100 or above.
Comparisons of ADI ratings of language compet-
ence also indicated significant differences between
the ‡70 and <70 bands (see Table 4b). All the indi-
viduals with a childhood IQ below 70 were rated as
having at least mild language problems as adults
and 15 out of the 23 had severe problems or no
speech. Nevertheless, over half (19) of the 70–99 IQ
group were rated as having moderate or severe
problems and although none of the 100+ IQ group
had failed to develop language, 3 out of the 10 were
still described as having severe problems.
There was no significant association between
childhood IQ and abnormal use of language in
adulthood. None of the individuals with an IQ of
100+ in childhood showed markedly abnormal use
of language in adulthood, and the majority (25/34)
of those in the 70–99 range also had only mild or
insignificant problems. However, although gener-
ally the frequency of severe problems area was
relatively low in each of the groups, the distribu-
tion of language abnormalities was variable (see
Table 4b).
The average number of ritualistic behaviours
reported by parents was somewhat higher in the
lower IQ band (falling in the Ômoderate rangeÕ) com-
pared to a rating in the ÔmildÕ range for the indi-
viduals with an IQ ‡ 70, but there was no significant
group difference (see Table 4b) and again, no differ-
ence between individuals with an IQ ‡ 100 and those
with an IQ of 70–99.
The relationship between early IQ and scores
on standardised tests was less consistent (see
Table 4b). Adult BPVS age equivalent scores were
significantly higher in the group with a childhood
IQ ‡ 70 but again, further examination of the data
indicated that there was no difference between
those with an IQ of 70–99 and those with an
IQ ‡ 100 (t ¼ .54, p ¼ .59). There was a significant
difference between sub-groups on Reading Com-
prehension, a marginal difference for Spelling and
no difference for Reading Accuracy. However,
Table 5 Child and adult characteristics of individuals with an initial IQ above 70 (full data missing on one individual)
Child (n)
Some Cognitive
Outcome speech decline
(n) IQ >100 at 5 PIQ >100 (>1sd) qualifications
Very good (7) None All 3 None
Good (7) 4 6 2 2
Fair (10) 4 8 2 2
Very/poor (20) 2 13 1 10
*Includes all individuals in independent, sheltered, voluntary or self-employed work. Employment status not known in one case.
Adult outcome of children with autism 221
whereas the majority (28 out of 45) in the ‡70 IQ
group scored above basal on all 3 attainment tests,
in the lower IQ band only 8 (35%) were able to score
above the basal of 5–6 years.
Overall, it appeared that performance IQ in
childhood was related to adult outcome in several
areas, especially those associated with social func-
tioning. The crucial cut-off point appeared to be
around 70, with very few individuals scoring below
this level as children achieving any real degree of
independence as adults. Nevertheless, even above
this cut-off, the picture proved to be very mixed. Of
the 45 individuals with an initial IQ of 70 or above,
only 7 could be considered to have a ÔVery GoodÕ
outcome, in terms of living reasonably independ-
ently, holding down a job, and having some outside
friendships. Another 7 were considered to have a
ÔGoodÕ outcome in that, although still living with
their families or in sheltered provision, they all had
some form of occupation and some outside friend-
ships. Ten were rated as ÔFairÕ. These individuals
were not able to live independently but they had
either sheltered employment or some form of
friendships and had some freedom to travel outside
their home or place of residence. The remainder was
significantly impaired, mostly living with their
families, in specialist autistic provision, or in insti-
tutional care. Individuals with a childhood IQ of
over 100 did not tend to have a better outcome than
those with an IQ between 70 and 99 and, in fact,
several individuals with an initial IQ of 70–99 had
made better progress (in terms of academic attain-
ments, jobs and general level of functioning) than
those with an IQ above 100.
Table 5 illustrates the very variable picture, ac-
cording to overall social outcome, amongst individ-
uals with an initial IQ of ‡70. Having an IQ well
within the normal range (i.e., 100+) as a child did not
seem to be related to adult outcome within this sub-
group. However, whereas all the individuals with a
ÔVery GoodÕ outcome, and the majority of those rated
ÔGoodÕ or ÔFairÕ, had some speech at 5, this was true
for only 13 out of 20 in the poorest outcome group.
Within the ÔVery/PoorÕ outcome group the majority
continued to show significant language impair-
ments, and as adults they had higher levels of ritu-
alistic behaviours. In addition, half of this group had
Adult (n)
Some
Formal Signif. language abnormal Mod-severe
In work* problems use of lang rituals
6 All None None None
5 All 1 4 2
2 3 3 5 4
1 None 18 16 12
(1 in past)
222 Patricia Howlin et al.

shown a decline in IQ from child to adulthood of at
least 1 standard deviation (15 points or more).
Verbal ability. When first seen, over 40% of the
group (29 children) failed to score at all on any
verbal test and 2 others had scores below 30.
Moreover, when the group was divided into those
with a childhood verbal IQ/IQ estimate of ‡70
(n ¼ 14) and those with a verbal IQ estimate of 30–
69 (n ¼ 23), no significant differences were found
on any of the social outcome variables (overall so-
cial competence, job, friendships, educational level
or residential status). There were no group differ-
ences in adulthood on frequency of ritualistic be-
haviours or abnormal language usage, and only a
weak effect for use of language in adulthood
(Mann–Whitney Z ¼ 2.19, p ¼ .028). There were,
however, a number of significant differences be-
tween the 31 individuals who were not able to
score on a language test to begin with and those
who were (see Tables 6a & b).
Similar comparisons were conducted taking lan-
guage level at age 5 as the independent variable.
Forty-two individuals were rated as having some
useful speech at 5 years (i.e., ADI scores 0–2) and
17 had little or no useful speech (ADI scores 3/No
speech). Early language data were not available for
the remaining cases. Compared with those who
had no speech at 5 years, those with speech had
significantly higher median scores for overall social
rating (7.0:8.5; Z ¼ 2.74, p ¼ .006) and residential
status (2.0:3.0; Z ¼ 3.17, p ¼ .002) but there were
no significant differences for educational level,
friendship rating, social or abnormal use of lan-
guage or ritualistic behaviour and only a marginal
difference for work level (Z ¼ 2.45, p ¼ .014). There
were no differences on reading or spelling tests.
There was a marginal difference on language
comprehension scores, as measured by the BPVS.
The average language age in adulthood of individ-
uals with some speech at age 5 was 9.7 years (sd
6.3) compared to 5.4 years (sd 4.9) for those
without useful speech at 5 (t ¼ 2.5, p ¼ .014).
Can we predict who will have the best outcomes?
Correlations between overall social rating scores and
childhood IQ measures, although significant, were
moderate. There was no significant difference in the
correlations between social outcome rating and child
verbal IQ or Performance IQ (r. social outcome/
VIQ ¼ .52; r. social outcome/PIQ ¼ .31; p ‡ 14) but
the failure of many children to obtain a verbal IQ
score in the early years limits the value of this vari-

Table 6a Childhood verbal scores and social outcome measures: differences between individuals scoring above 30 IQ on verbal test
as children and those below this level, or unable to score

Mean (sd)

Child language test VIQ ‡ 30 (n ¼ 37) VIQ <30/not testable (n ¼ 31)

b
Parental rating Mann Whitney Z a p
Residential status 2.69 (1.43) 3.74 (1.93) 2.65 .008
Quality of friendships 1.89 (1.35) 2.29 (1.13) 1.18 .236
Educational level 3.38 (1.19) 3.50 (1.30) 1.10 .271
Level of work 2.03 (1.18) 2.68 (.83) 2.79 .005
Total social outcome rating 6.44 (2.95) 8.13 (3.04) 2.68 .007
a
Means, rather than medians, are presented for ease of reference. However, because of the nature of the data, non-parametric
statistics were used for group comparisons.
b
The higher the score the greater the abnormality, see Appendix 2.

Table 6b Childhood verbal scores and adult outcome measures: Differences between individuals scoring above 30 IQ on verbal test
as children and those below this level, or unable to score

Mean (sd)

Child language test VIQ ‡ 30 (n ¼ 37) VIQ <30/not testable (n ¼ 31)

b
Parental rating Mann Whitney Z a p
Social use of language 8.03 (6.44) 14.03 (7.29) 3.74 .001
Abnormal use of language 2.79 (2.94) 3.20 (2.31) 0.95 .343
Ritualistic behaviours 5.22 (4.11) 6.93 (4.67) 1.82 .069
Standard test scores Mean in years (sd) Mean in years (sd) t
BPVS age equivalent 10.6 (5.9) 5.5 (5.4) 3.69 <.001
Neale accuracy age (yrs) 11.1 (1.7) 9.4 (2.4) 2.64 .012
Neale comprehension age (yrs) 8.7 (2.4) 8.8 (2.4) 0.16 .875
Spelling age (yrs) 11.8 (2.5) 11.2 (3.9) 0.66 .512
a
Means for rating scores, rather than medians, are presented for ease of reference. However, because of the nature of the data, non-
parametric statistics were used for group comparisons.
b
The higher the score the greater the abnormality, see Appendix 2.
 Adult outcome of children with autism 223

able in predicting outcome. Moreover, as is evident eral predictive value. More individuals (n ¼ 11) fell
from Figure 4, having a higher verbal IQ in childhood within the PIQ ‡ 70 plus VIQ ‡ 70 range, and of
did not necessarily predict a positive outcome in these 5 had a ÔGoodÕ/ÔVery GoodÕ outcome; only 2
adulthood. Only 6 individuals out of the 14 with an were rated ÔPoorÕ. Within the group PIQ ‡ 70/
initial VIQ ‡ 70 had a ÔGoodÕ/ÔVery GoodÕ outcome VIQ ‡ 50 (n ¼ 20), 8 had a ÔGoodÕ/ÔVery GoodÕ out-
and 4 were rated as ÔPoorÕ. Of the 26 individuals who come; 7 were rated as ÔFairÕ and 5 as ÔPoorÕ. (See
obtained a verbal IQ ‡ 50 as children, 9 achieved a Figure 4.)
ÔGoodÕ/ÔVery GoodÕ outcome; 9 others were rated as In some ways it proved easier to identify corre-
ÔFairÕ and 8 as ÔPoor/Very PoorÕ. The proportion rated lates of ÔPoorÕ outcome than the variables predictive
as ÔGoodÕ/ÔFairÕ/ÔPoorÕ was also similar for individu- of good prognosis. No one with a childhood per-
als with an initial verbal IQ below 50. And, as pre- formance IQ below 70 and a verbal IQ below 30
viously noted, even amongst those who obtained a achieved even a ÔFairÕ rating in adulthood (see
performance IQ score ‡ 70 as children, less than Figure 5) and only one individual with PIQ
one-third (14/44) was rated as having a ÔGoodÕ out- <70 + VIQ < 50 did so. Most individuals with an
come in adulthood (see Figure 4). initial performance IQ below 70 (n ¼ 23) were highly
Whilst neither Performance nor Verbal IQ alone in dependent as adults; only one had a ÔVery GoodÕ
childhood proved a particularly good predictor of outcome and 3 others were rated as ÔFairÕ. Having a
later outcome, the combination of these measures low verbal IQ initially was also associated with
did provide a more reliable indicator. All those with a poorer outcome, but because VIQ increased sub-
verbal IQ 70 and a Performance IQ ‡ 100 had a stantially in some individuals over time the rela-
ÔGoodÕ or ÔFairÕ outcome, but as only 3 cases fell tionship was less consistent. Thus, although among
within this grouping this is likely to be of little gen- those with a childhood VIQ <70 (n ¼ 54) 35 had a

9 10 14 1
6 5 8
No.achieving social outcome rating:

11 10 V/Good
9
4 7 Fair
2 4 V/Poor
15 20
4 8 5
2
VI VI VI PI PI PI PI
Q> Q> Q> Q> Q> Q& Q>
70 50 30 70 10 VI 70
* * * 0/V Q> /V
IQ 70 IQ
>7 * >5
0* 0

Cognitive level as children (*outcome data missing for 1case)

Figure 4 Variables associated with good adult outcome

35 No. achieving social rating:

19 31 24
12 16 V/Poor
Fair
9 V/Good
4 2
3 9 6 5
1 1
PIQ<70, PIQ<70; PIQ<70 VIQ<70* VIQ<50 VIQ<30
VIQ<30 VIQ<50

Cognitive level as children (* data missing on one case)

Figure 5 Variables associated with poor adult outcome

224 Patricia Howlin et al.
ÔPoorÕ/ÔVery PoorÕ outcome, 4 individuals were rated
as ÔVery GoodÕ, 5 as ÔGoodÕ and 9 as ÔFairÕ as adults
(data missing on one case). Of the 41 with a verbal
IQ <50 in childhood, the majority (31) had ÔVeryÕ/
ÔPoorÕ outcomes but 2 were rated as ÔVery GoodÕ, 4
as ÔGoodÕ and 4 as ÔFairÕ. Even amongst those who,
as children, were untestable or had a verbal IQ <30
(n ¼ 31), 2 individuals were rated as ÔVery GoodÕ, 3
as ÔGoodÕ and 2 as ÔFairÕ in adulthood. Thus despite
a relatively strong statistical correlation between
early verbal IQ scores, the clinical value of this
variable in predicting individual outcomes was
limited.
Gender differences in outcome
As there were only 7 women in the sample, any
conclusions about possible gender differences can
only be tentative. Males and females were a similar
age when first assessed (M:7.1 yrs; F:7.6 yrs) and
their non-verbal IQ scores were comparable (M:80.2
yrs; F:79.6 yrs). At follow-up women were slightly
older (M:28.8 yrs; F:34.0 yrs) but the difference
was not significant. Scores for verbal and non-ver-
bal IQ, and on language, reading and spelling tests
did not differ. As in the total sample, Performance
IQ scores remained very stable over time and only
one woman showed a major change (from an initial
IQ of 137 to a follow-up IQ of 74); the remainder all
remained within the same band as when first as-
sessed. Verbal IQ scores were somewhat more
variable. Two women who were initially untestable
on a verbal IQ test scored above 70 at follow-up; 3
moved upwards (from VIQ £ 49 to VIQ 50–69); one
individual who had scored above 70 as a child
could no longer score on a verbal IQ test in adult-
hood and one remained untestable. There were no
differences between men and women on measures
of language level, abnormal use of language or re-
petitive and stereotyped behaviours, or on overall
social outcome ratings (Mann–Whitney U test Z
values all p ‡ .05). However, no female was rated as
having a ÔGoodÕ outcome in adulthood and five were
within the ÔPoorÕ or ÔVery PoorÕ groups. None had
ever attended normal school or obtained any formal
qualifications. Only one had a job of any sort
(helping out in the family firm); the remainder was
all in special day or residential centres. None was
reported to have friends or to be living independ-
ently; 3 were in some form of residential unit, one
lived with her parents and 3 were in long-stay
hospitals.
Discussion
General functioning in adulthood
This report on 68 individuals with autism and a non-
verbal IQ of at least 50 represents one of the largest
systematic follow-up studies of this condition in
adult life. Other relatively large follow-up studies
(Lockyer & Rutter, 1969, 1970; Rutter & Lockyer,
1967; Rutter et al., 1967; Kobayashi et al., 1992)
have included both adults and children/adolescents
and have combined data on those who are severely
intellectually impaired (and for whom prognosis is
almost certain to be poor) with those of higher IQ.
The outcome measures used in the present study
also cover many different domains, and were based
on information from standardised tests and struc-
tured interviews of established reliability and valid-
ity. Detailed and systematic IQ data, rather than
estimates of cognitive functioning, are also available
from child to adulthood.
In terms of general outcome, the results sugges-
ted that many people had continued to make pro-
gress since they were seen in childhood. One-fifth
had managed to obtain some academic qualifica-
tions at school and 5 of these individuals had gone
on to college or university, with 2 studying at
postgraduate level. Almost a third was in employ-
ment, and around a quarter of the group was de-
scribed by parents as having some friendships
involving shared interests or activities. Neverthe-
less, only 8 individuals could be rated as having
achieved a high level of independence. Six of these
were in paid employment and all showed a fair de-
gree of independence, although only 2 (both of
whose mothers had died) survived with little regular
support (financial, social and/or emotional) from
their families.
The majority, however, remained highly dependent
on their families or required some form of residential
provision. Even amongst those in employment, jobs
tended to be poorly paid and did not provide indi-
viduals with adequate financial support to live in-
dependently. One young man, for example, who had
become a skilled charcoal burner, remained in a
MENCAP hostel. Although he would have preferred
to live independently, he could not afford to do so.
Another man, working in a graveyard, wanted to
move to Social Services accommodation, rather than
living at home. As his wages could in no way cover
the costs of this, he was given the choice of keeping
his job and remaining at home, or leaving home and
giving up his job.
Overall, few individuals could be considered to be
fully self-supporting and the proportion (57%) of
those rated as having a ÔPoorÕ or ÔVery PoorÕ outcome
remains similar to that reported in original follow-up
study of Rutter and his colleagues (1967) over 30
years ago. Follow-up studies in Sweden (Gillberg &
Steffenburg et al., 1987), Japan (Kobayashi et al.,
1992) and North America (Ballaban-Gil et al., 1996;
Szatmari et al., 1989; Venter et al., 1992) indicate
similarly high levels of dependence, low economic
status, and persisting language and/or behavioural
difficulties. There are, of course, some positive find-
ings. The percentage in open or sheltered employ-
ment, although relatively low (34%), is much higher

than the figures of below 10% reported in studies
conducted in the 1970s and 80s. And, whereas al-
most half of the individuals in the Rutter and Lock-
yer studies were living in long-stay hospitals as
adults, this was the case for only a minority in the
present study. The proportion of individuals able to
live semi/dependently, whilst small (around 10%), is
also higher than the very low rates reported in some
earlier studies (Ballaban-Gil et al., 1996; Gillberg
et al., 1987; Rutter et al., 1967; Kobayashi et al.,
1992; Lotter, 1974).
However, despite the group’s relatively high IQ,
attainments in many areas were disappointingly low.
Not only had very few obtained any formal qualifi-
cations at school, even basic skills were very poor
and almost half the group was unable to score above
basal level (of 5 and 6 years respectively) on either
the spelling or reading tests. Similarly, their mean
age equivalent on the British Picture Vocabulary
Scale was only eight years.
On the whole, it would appear that the huge in-
crease in educational facilities for children with
autism over the past 3 decades has not necessarily
resulted in significant improvements in outcome for
adults. What then, if anything, is predictive of out-
come? All previous major follow-up studies of autism
have indicated that outcome is almost invariably
poor for individuals with a childhood IQ of below 50
(cf. Nordin & Gillberg, 1998; Lord & Bailey, 2002).
However, as noted earlier, few people of this cognitive
level, whatever the cause of their difficulties, achieve
independence as adults. The present study therefore
focused on prognosis in individuals with an IQ
above 50.
The first issue investigated concerned the con-
sistency of IQ over time. It has often been suggested
that the results of early testing with children with
autism are likely to be unreliable because of the
children’s poor understanding or social withdrawal.
However, if the appropriate tests are chosen, it
should not prove difficult to obtain good cooperation
and reliable results (Clark & Rutter, 1979, 1981;
Lord & Bailey, 2002). In the present study it was
apparent that IQ remained remarkably stable over
time. Correlations between adult and childhood IQs
were highly significant; in many cases actual IQ
scores showed little change and there were few sig-
nificant differences in overall measures of either
performance or verbal IQ over time. Over three-
quarters (35/45) of those with an initial IQ of 70 or
above remained within this range. There was also a
significant relationship between adult IQ and other
outcome measures.
Other follow-up studies have also noted the relat-
ive stability of IQ in autism, although generally over
shorter time periods (cf. Ballaban-Gil et al., 1996;
Lockyer & Rutter, 1970; Venter et al., 1992). The
majority (51/68) in the present sample was beyond
the age of 5 when first seen, and the consistency in
IQ here may be a function of the age at which they
Adult outcome of children with autism 225
were initially assessed. Thus the children tested be-
fore 5 years showed greater change in performance
IQ than those assessed after 5 (a mean decrease of
13 points compared to a mean drop of 2.5 points in
the older group; t ¼ 1.96, df 66, p ¼ .054). However,
this is in marked contrast to the significant im-
provements in IQ (of 30 points or more) reported for
children aged between 2 and 4 years who have been
enrolled in early intensive, behaviourally based pre-
school programmes (Lovaas, 1987, 1993; McEachin,
Smith, & Lovaas, 1993). All that can be concluded is
that, if children are tested in the early school years,
and if appropriate tests are used, then, in the ab-
sence of intensive, specialised intervention, marked
changes in IQ from child to adulthood are infrequent.
Intensive therapy might have well have resulted in
greater changes, although it is also important to note
that serious concerns have been raised about the
way in which IQ changes have been calculated in
these early intervention studies (Magiati & Howlin,
2001).
Given that early IQ testing does produce stable
findings, the next issue concerns the relationship of
childhood IQ to later outcome. Lord and Bailey
(2002) suggested that variations in non-verbal IQ
between 50 and 70 have a somewhat similar effect to
those below 50 (although the effects are less), but
that variations in non-verbal IQ within the normal
range are of little predictive value. Our results are
broadly in keeping with their conclusion. As com-
pared to individuals with an initial IQ over 70, those
with an initial performance IQ in the range of 50–69
appeared to have a much poorer prognosis in
adulthood. The majority in the latter group remained
highly dependent, and few had jobs or friends of any
kind. Language skills were also generally very im-
paired. Outcome for those in the 70+ IQ range was
significantly better in almost every domain assessed.
These findings are important, for whilst previous
studies had indicated that an IQ cut-off of around 50
is critical in determining which individuals do and do
not do well, here it appears that only individuals with
an IQ in the normal range (70+) have a real chance of
living independently as they reach adulthood. Below
this level outcome is likely to be much poorer.
Nevertheless, the findings also showed that, al-
though having an IQ of at least 70 is a crucial
prognostic factor, above this level outcome can still
be very variable. There was little difference in
adulthood between those with a childhood IQ of 100
or more, compared with those with an IQ between 70
and 99, and even those in the highest IQ group ex-
perienced many problems as adults. One individual,
for example, with a childhood IQ of 119, and high
ratings on many of the adult outcome measures, was
frequently in trouble, at work and elsewhere, for in-
appropriately following or touching females, and he
appeared to have little or no understanding of the
potential impact of these behaviours. Another indi-
vidual (initial IQ 80) had obtained a degree and
226 Patricia Howlin et al.
various computing qualifications in adulthood. He
had also married (although he later divorced). How-
ever, he had no ability to perceive othersÕ social in-
tentions and was constantly being defrauded of large
sums of money because of this. A third individual
(initial IQ 97) was in danger of losing his job on
several occasions because of his inability to under-
stand social rules. Despite working in a culturally
mixed office he often gave unwitting offence by re-
ferring disparagingly to people’s race or colour or to
the fact that they were ÔforeignersÕ.
There were also certain aspects of adult function-
ing, mainly related to autistic-type behaviours (rou-
tines, rituals, stereotypies etc.), that did not appear
to be strongly related to early IQ. Although the more
able group was less likely to show very severe diffi-
culties in these areas, the distribution of such
symptoms was generally fairly evenly spread and
within each IQ band, the majority of individuals
continued to have at least mild to moderate problems
associated with repetitive and stereotyped behav-
iours. Thus, although having a childhood IQ within
the normal range is clearly a crucial factor in deter-
mining outcome, within this sub-group the presence
of ritualistic and/or stereotyped behaviours may still
prevent individuals from attaining an optimal level of
functioning. In the case of one man (initial IQ 80), for
example, ritualistic checking behaviours seriously
compromised his ability to work independently, de-
spite his having obtained various postgraduate
qualifications. In another (initial IQ 114), the level of
anxiety engendered by his need for predictability
meant that he could only cope within highly sup-
portive, specialist autism provision as an adult. The
very fixed routines and preoccupations of another
(initial IQ 107) also resulted in his needing specialist
support in adulthood, both in work and accommo-
dation.
It is clear that childhood performance on non-
verbal tests of intelligence, whilst being a relatively
good predictor of outcome, is by no means perfect
and Lord and Bailey (2002) have proposed that
childhood verbal IQ is a far more reliable indicator of
later functioning. However, as is apparent from the
present study, many children who failed to score
above the basal level (or even to score at all) on verbal
IQ tests when young went on to show significant
improvements in this area over time. In the case of
other children, who were able to obtain a verbal IQ
score when first assessed, the relationship with
adult outcome was very variable. Thus, from a clin-
ical perspective, verbal IQ in early childhood appears
to have only limited predictive value.
Lord and Bailey (2002) also suggested that the
presence of useful speech by age 5 is highly predic-
tive of later outcome. Certainly, for many young
children it is much easier to obtain information of
this kind than to obtain a verbal IQ score, although
there may be some problems of recall if interviewing
parents of older individuals. However, in the present
study this variable, too, was only weakly associated
with adult outcome. Over 40% of children who had
little or no language when first assessed had devel-
oped useful language as adults, and the higher their
linguistic levels as adults, the more likely they were
to do well on a range of other outcome measures.
Other research has pointed to the impact that im-
provements in language may have on the develop-
mental trajectory of children with autism (Szatmari,
2000) but as yet we have little information on what is
associated with such improvement. In an ongoing
study we are examining in more detail patterns of
early cognitive functioning that may be associated
with increases, or decline, in levels of functioning,
and the ways in which these may contribute to the
picture in adult life.
Summary
There is no doubt that therapeutic and educational
provision for individuals with autism has improved
over the past 3 decades, and, as this and other
studies show, a substantial minority of adults, al-
though continuing to be affected by their autism, can
find work, may live independently, and develop
meaningful social relationships with others. Never-
theless, the majority remains very dependent on
parents or others for support.
One of the most significant factors determining
outcome appears to be level of intellectual function-
ing in childhood, and IQ scores, at least on non-
verbal tests, tend to remain relatively stable over
time. However, whereas earlier studies had sugges-
ted that an IQ level of above 50 was associated with a
more positive outcome, the findings here indicate
that only individuals with an IQ of over 70 are likely
to do well. Nevertheless, even within this higher
functioning group, outcome tends to be very variable
and it seems that the fundamental deficits associ-
ated with autism, in particular the degree of ritual-
istic and stereotyped behaviours, may at times
ÔswampÕ the effects of a relatively high IQ (see also
Howlin et al., 2000; Mawhood et al., 2000). In addi-
tion, despite the relatively small changes in IQ
overall, in a minority of cases there were consider-
able changes in IQ levels (both upwards and down-
wards) over time. Factors influencing change from
child to adulthood, and the trajectories that different
individuals may follow, are clearly important issues
for future research.
Finally, the ability to function adequately in
adulthood life may depend as much on the degree of
support offered (by families, employment and social
services) as on basic intelligence (Lord & Venter,
1992; Mawhood & Howlin, 1999). Although admis-
sions to hospital care have fallen, and expectations
about the future for people with disabilities generally
have risen over the years, dedicated services for
adults with autism would not seem to have kept pace

with the growth in specialist educational provision
for children. A focus on access to more extensive and
appropriate supported living and employment
schemes could help to ensure much greater progress
in the future.
Correspondence to
Patricia Howlin, Department of Psychology, St Geor-
ge’s Hospital Medical School, Cranmer Terrace, Lon-
don SW17 0RE, UK; Email: phowlin@sghms.ac.uk
References
Ballaban-Gil, K., Rapin, I., Tuchman, R., & Shinnar, S.
(1996). Longitudinal examination of the behavioral,
language, and social changes in a population of
adolescents and young adults with autistic disorder.
Pediatric Neurology, 15, 217–223.
Bolton, P., Macdonald, H., Pickles, A., Rios, P., Goode,
S., Crowson, M., Bailey, A., & Rutter, M. (1994). A
case-control family history study of autism. Journal of
Child Psychology and Psychiatry, 35, 877–900.
Clark, P., & Rutter, M. (1979). Task difficulty and task
performance in autistic children. Journal of Child
Psychology and Psychiatry, 20, 271–285.
Clark, P., & Rutter, M. (1981). Autistic children’s
responses to structure and to interpersonal demands.
Journal of Autism and Developmental Disorders, 11,
201–217.
Creak, M. (1963). Childhood psychosis: A review of 100
cases. British Journal of Psychiatry, 109, 84–89.
Dunn, L., Dunn, L., Whetton, C., & Pintillie, D. (1982).
British Picture Vocabulary Scale. Slough, Bucks:
NFER-Nelson Publishing Co.
Dunn, L.M. (1965). Peabody Picture Vocabulary Test.
Circle Pines, MN: American Guidance Service.
Eisenberg, L. (1956). The autistic child in adolescence.
American Journal of Psychiatry, 112, 607–612.
Gillberg, C., & Steffenburg, S. (1987). Outcome and
prognostic factors in infantile autism and similar
conditions: A population-based study of 46 cases
followed through puberty. Journal of Autism and
Developmental Disorders, 17, 272–288.
Howlin, P., & Goode, S. (1998). Outcome in adult life for
individuals with autism. In F. Volkmar (Ed.), Autism
and developmental disorders. New York: Cambridge
University Press.
Howlin, P., Mawhood, L.M., & Rutter, M. (2000). Autism
and developmental receptive language disorder: A
follow-up comparison in early adult life: II. Social,
behavioural and psychiatric outcomes. Journal of
Child Psychology and Psychiatry, 41, 561–578.
Hutton, J. (1998). Cognitive decline and new problems
arising in association with autism. Unpublished doc-
tor of clinical psychology thesis, University of London.
Kobayashi, R., Murata, T., & Yashinaga, K. (1992). A
follow-up study of 201 children with autism in
Kyushu and Yamaguchi, Japan. Journal of Autism
and Developmental Disorders, 22, 395–411.
LeCouteur,A.,Rutter,M.,Lord,C.,Rios,P.,Robertson,S.,
Hografer, M., & McLennan, M. (1989). Autism
Adult outcome of children with autism 227
Diagnostic Interview: A standardised, investigator
based instrument. Journal of Autism and Develop-
mental Disorders, 19, 363–387.
Levine, M.N. (1982). Leiter International Performance
Scale: A handbook. Los Angeles, CA: Western
Psychological Services.
Lockyer, L., & Rutter, M. (1969). A five-to-fifteen year
follow-up study of infantile psychosis: III. Psycholo-
gical aspects. British Journal of Psychiatry, 115,
865–882.
Lockyer, L., & Rutter, M. (1970). A five-to-fifteen year
follow-up study of infantile psychosis: IV. Patterns of
cognitive ability. British Journal of Social and Clinical
Psychology, 9, 152–163.
Lord, C., & Bailey, A. (2002). Autism spectrum dis-
orders. In M. Rutter & E. Taylor (Eds.), Child and
adolescent psychiatry (4th edn, pp. 664–681). Oxford:
Blackwell Scientific.
Lord, C., Rutter, M., & Le Couteur, A. (1994). Autism
Diagnostic Interview-Revised: A revised version of a
diagnostic interview for carers of individuals with
possible pervasive developmental disorders. Journal
of Autism and Developmental Disorders, 24, 659–685.
Lord, C., & Schopler, E. (1985). Differences in sex ratios
in autism as a function of measured intelligence.
Journal of Autism and Development Disorders, 15,
185–193.
Lord, C., & Venter, A. (1992). Outcome and follow-up
studies of high functioning autistic individuals. In
E. Schopler & G.B. Mesibov (Eds.), High functioning
individuals with autism (pp. 187–200). New York:
Plenum.
Lotter, V. (1974a). Factors related to outcome in autistic
children. Journal of Autism and Childhood Schizo-
phrenia, 4, 263–277.
Lotter, V. (1974b). Social adjustment and placement of
autistic children in Middlesex: A follow-up study.
Journal of Autism and Childhood Schizophrenia, 4,
11–32.
Lovaas, O.I. (1987). Behavioral treatment and normal
educational and intellectual functioning in young
autistic children. Journal of Consulting and Clinical
Psychology, 55, 3–9.
Lovaas, O.I. (1993). The development of a treatment–
research project for developmentally disabled and
autistic children. Journal of Applied Behavior Analy-
sis, 26, 617–630.
Magiati, I., & Howlin, P. (2001). Evaluating the long-
term effectiveness of early intervention for young
children with autism: Preliminary baseline data and
methodological difficulties. Autism: International
Journal of Research and Practice, 5, 407–429.
Mawhood, L.M., Howlin, P., & Rutter, M. (2000). Autism
and developmental receptive language disorder: A
follow-up comparison in early adult life: I. Cognitive
and language outcomes. Journal of Child Psychology
and Psychiatry, 41, 547–559.
Mawhood, L.M., & Howlin, P. (1999). The outcome of a
supported employment scheme for high functioning
adults with autism or Asperger syndrome. Autism:
International Journal of Research and Practice, 3,
229–253.
McEachin, J.J., Smith, T., & Lovaas, O.I. (1993). Long-
term outcome for children with autism who received
228 Patricia Howlin et al.
early intensive behavioral treatment. American Jour-
nal of Mental Retardation, 97, 359–372.
Mecham, M.J. (1958). Verbal Language Development
Scale. Los Angeles, CA: Western Psychological Servi-
ces.
Neale, M.D. (1977). Neale analysis of reading ability
(2nd edn, reprinted 1997). London: Macmillan.
Nordin, V., & Gillberg, C. (1998). The long-term course
of autistic disorders: Update on follow-up studies.
Acta Psychiatrica Scandinavica, 97, 99–108.
Raven, J. (1976). Standard progressive matrices. Lon-
don: H.K. Lewis.
Reynell, J. (1977). Reynell Developmental Language
Scales. Oxford: National Foundation for Educational
Research.
Rumsey, J.M., Rapoport, J.L., & Sceery, W.R. (1985).
Autistic children as adults: Psychiatric social and
behavioral outcomes. Journal of the American Acad-
emy of Child Psychiatry, 24, 465–473.
Rutter, M. (1966). Behavioural and cognitive character-
istics of a series of psychotic children. In J. Wing
(Ed.), Early childhood autism (pp. 51–81). London:
Pergamon.
Rutter, M. (1972). Childhood schizophrenia reconsid-
ered. Journal of Autism and Childhood Schizophrenia,
2, 315–337.
Rutter, M. (1978). Diagnosis and definition. In M. Rutter
& E. Schopler (Eds.), Autism: A reappraisal of
concepts and treatment (pp. 1–27). New York: Plenum.
Rutter, M., & Lockyer, L. (1967). A five to fifteen year
follow-up study of infantile psychosis: I. Description
of the sample. British Journal of Psychiatry, 113,
1169–1182.
Rutter, M., Greenfeld, D., & Lockyer, L. (1967). A five to
fifteen year follow-up of infantile psychosis: II. Social
and behavioural outcome. British Journal of Psychi-
atry, 113, 1183–1189.
Schonell, F., & Schonell, F. (1960). Diagnostic and
attainment testing. Edinburgh: Oliver and Boyd.
Sparrow, S., Balla, D., & Cicchetti, D. (1984). Vineland
Adaptive Behavior Scales. Circle Pines, MN: American
Guidance Service.
Stutsman, R. (1948). Merrill Palmer Scale of Mental
Tests. Los Angeles, CA: Western Psychological Servi-
ces.
Szatmari, P. (2000). Perspectives on the classification
of Asperger syndrome. In A. Klin, F.R. Volkmar, &
S.S. Sparrow (Eds.), Asperger syndrome (pp. 403–
417). New York: Guildford Press.
Szatmari, P., Bartolucci, G., Bremner, R.S., Bond, S., &
Rich, S. (1989). A follow-up study of high functioning
autistic children. Journal of Autism and Developmen-
tal Disorders, 19, 213–226.
Tantam, D. (1991). Asperger’s syndrome in adulthood.
In U. Frith (Ed.), Autism and Asperger syndrome
(pp. 147–183). Cambridge: Cambridge University
Press.
Terman, L.M., & Merrill, M.A. (1961). Stanford Binet
Intelligence Scale (3rd rev.). London: Harrap.
Venter, A., Lord, C., & Schopler, E. (1992). A follow-up
study of high functioning autistic children. Journal of
Child Psychology and Psychiatry, 33, 489–507.
Wechsler, D. (1974). Wechsler Intelligence Scale for
Children–Revised. New York: Psychological Corpora-
tion.
Wechsler, D. (1981). Wechsler Adult Intelligence Scale–
Revised. New York: Psychological Corporation.
Wechsler, D. (1990). Wechsler Pre-school and Primary
Scale of Intelligence: Revised UK edition (WPPSI-R UK).
Sidcup, Kent: The Psychological Corporation.
Manuscript accepted 17 December 2002
Appendix 1
Selection of subjects for adult follow-up study
Inclusion criteria
These were the same as for the genetic family study
of Bolton et al. (1994), i.e.:
Singleton (not twin)
• Caucasian
• No evidence of specific medical disorders possibly
related to autism (tuberous sclerosis, phenyl-
ketonuria, infant spasms, neurofibromatosis)
• No evidence of major physical or sensory impair-
ment
• No prolonged period in institutional care.
(See Bolton et al., 1994, for further details.) How-
ever, in contrast to the Bolton et al. (1994) sample:
• Adoptees were included, as were individuals later
found to have chromosomal abnormalities.
Additional criteria for inclusion in the present
study:
• First diagnostic assessment prior to the age of 16
years (individuals initially referred before this age
because of primary concerns about mental health
problems/deterioration were not included)
• Minimum gap of 3 years between initial and follow-
up assessments
• Initial performance IQ P50.
Appendix 2
Details of assessments
Ratings of adult functioning
These were based principally on parental informa-
tion from the Autism Diagnostic Interview (ADI;
Le Couteur et al., 1989). This is the earlier version of
the revised ADI (ADI-R; Lord, Rutter, & Le Couteur,
1994). For further details of scoring criteria see
Le Couteur et al., 1989.
Social functioning and independence
Work ratings were 0: employed or self-employed;
1: voluntary work/job training or low-pay scheme;
2: supported/sheltered employment; 3: in special
centre/no occupation
Friendship. ÔFrequencyÕ of friendships was rated
according to the numbers of friends/acquaintances

the individual had, and ÔQualityÕ according to the
degree of sharing and mutual participation the re-
lationship involved. Scores ranged from 0 (>1 close
friendships involving sharing and exchange of con-
fidences and range of different activities together) to
3 (no friends; no joint activities).
Independence ratings were 0: living independently;
1: in semi-sheltered accommodation (or still at home)
but with high degree of autonomy; 2: living with
parents, some limited autonomy; 3: in residential
accommodation with some limited autonomy; 4:
specialist autistic or other residential accommoda-
tion with little or no autonomy, 5: in hospital care or
at home because nowhere else would accept the in-
dividual.
A composite measure of overall social functioning
was based on the sum of scores in the areas noted
above. Ratings were as follows:
0 ¼ ÔVery GoodÕ outcome – i.e., achieving a high level
of independence, having some friends and a job
(total for all 3 areas above 0–2)
1 ¼ ÔGoodÕ outcome – generally in work but requiring
some degree of support in daily living; some
friends/acquaintances (total 3–4)
2 ¼ ÔFairÕ – has some degree of independence, and
although requires support and supervision does
not need specialist residential provision; no
close friends but some acquaintances (total 5–7)
3 ¼ ÔPoorÕ – requiring special residential provision/
high level of support; no friends outside of
residence (total 8–10)
4 ¼ ÔVery PoorÕ – needing high-level hospital care, no
friends; no autonomy (total 11).
Composite ratings were checked both by PH & SG
and agreement for all cases was 100%.
Current language usage was assessed using a
composite score obtained from the ADI. The com-
posite was based on the sum of ratings for utterance
length, spontaneous communication, conversational
ability, reporting of events, amount of social com-
munication, and intonation and vocal expressive-
ness. Scores for each variable were as follows:
0 ¼ little or no abnormality; 1 ¼ mild problems (not
interfering with intelligibility of speech, etc.);
2 ¼ difficulties that occurring frequently or sufficient
to interfere with intelligibility; 3 ¼ problems both
frequent and severe. Subjects with no intelligible
speech were assigned a score of 8. (See Le Couteur
et al., 1989, for further details.)
A composite score of 0 indicated normal or near-
normal linguistic functioning (total 0–2); a score of 1
mild to moderate impairments (total 3 to 7); a score of
2 moderate impairment (8 to 10); a score of 3 severe
impairment (11+), and a score of 4 no language.
Abnormal use of language was based on the total
ADI ratings for current immediate echolalia, stereo-
typed utterances, pronoun reversal, neologisms,
idiosyncratic language and inappropriate speech.
For each variable a score of 0 indicated little or no
Adult outcome of children with autism 229
abnormality; 1 ¼ mild/occasional problems; 2 ¼ fre-
quent occurrence but with some appropriate use of
speech; 3 ¼ almost all speech abnormal. Subjects
with no intelligible speech were not rated on these
variables.
A composite score of 0 indicated little or no
abnormal usage (total 0–1); a score of 1 occasional or
mild abnormalities (total 2 to 5); a score of 2 definite
and frequent abnormalities (total 6+). The maximum
obtained by any individual on this measure was 10.
Subjects with little or no useful language were not
rated on this measure.
Autistic-type problems in adulthood
Ratings were based on the ADI scores for presence of
verbal rituals, unusual preoccupations, rituals and
compulsions, resistance to change, unusual attach-
ment to objects, unusual sensory interests, and
complex hand/body mannerisms. For each variable
a score of 0 indicated little or no abnormality;
1 ¼ mild/occasional problems; 2 ¼ frequent prob-
lems; 3 ¼ frequent problems with considerable dis-
ruption to family life.
A composite score of 0 indicated few or no prob-
lems (total 0–1); a score of 1 occasional or mild
problems (total 2 to 5); a score of 2 definite and fre-
quent problems in several areas (total 6 to 10) and a
score of 3 severe problems in multiple areas (total
score 11+).
Cognitive testing
A Ôbest estimateÕ adult IQ was assigned to each
subject for the purpose of banding into IQ groups.
The IQ score was assigned in a hierarchical manner,
taking either the WAIS Full Scale or Performance IQ
score (whichever was higher) if available (n ¼ 46).
One individual scored only on the WAIS Verbal Scale
and his Ôbest estimate IQÕ was based on this. If
subjects could not score on the WAIS, the Raven’s IQ
score was used (n ¼ 15); if not then the Leiter
(n ¼ 5), then the Merrill Palmer (n ¼ 1).
Language – BPVS scores
In the majority of cases, raw scores were below
basal for the individual’s age (n ¼ 40) or chronolo-
gical ages were above ceiling (n ¼ 7) and it hence
was not possible to calculate a standard language
score; instead language age equivalents were used.
The 7 individuals with raw scores above the level for
which an age score could be calculated from the test
norms were randomly assigned scores between 19
years 6 months and 21 years (i.e., between the
maximum test age and the upper confidence limit
for this). The 4 who scored below the minimum age
level of 1 year 8 months were randomly assigned
scores between zero and 1 year 8 months (see
Mawhood et al., 2000).